Before we get going, here’s something you’ll want to know about:
The AI + CRE Quickstart Course is now live.
If you’ve been sitting on the sidelines, waiting for the right time to start putting AI to work in CRE, this is how you get going in a hurry.
The most actionable tools, tutorials, and prompts for CRE, packaged to let you hit the ground running.
If you’re ready to start saving hours every week by leveraging AI, grab your 7 day free trial to the Studio below.
OpenAI Codex Adds Record & Replay Feature: This new feature allows you to record yourself doing a repetitive task on your Mac while Codex watches, and it writes itself a plain-English skill describing what you got done. Next time, it uses your computer to run the job itself, actually clicking the buttons, dragging the files, typing into the fields, and moving between apps the way you did, but with the new inputs.
CoStar Launches Apartments.com AI, Replacing Dropdown Filters With Conversational Search: Renters can now type or speak what they want and AI searches Apartments.com availabilities to provide the user matching availabilities. CoStar reports 56% more leads versus traditional search.
Data Centers Get a Grid Connection Fast Track, While Studies Say Home-Price Decline Fears from Data Centers are Overblown: Last week the Federal Energy Regulatory Commission ordered six major grid operators to fast-track data-center grid connection, after the Energy Secretary warned that slow grid connections were threatening U.S. competitiveness in AI. Notably, in the same week, two studies (George Mason and Integra Realty Resources) found data centers don't adversely affect nearby home values.
If you’re going down the AI rabbit hole, you’ve probably run into the concept of AI skills, and Claude skills specifically, since Claude has dominated the AI conversation in 2026.
If you don't know what a skill is yet, it's a set of instructions you package up so your AI tool can run a task, or even a multi-step process, just by calling the skill. They aren't just a Claude-only thing anymore either. They work in Claude, the business tiers of ChatGPT, and Google Gemini’s developer tools. A skill can build documents from templates, find and read files you point it towards, tap into your other tools through MCP connectors, and quite a bit more.
At CRE AI Studio we're huge advocates of learning to build and use skills. Just last week we released our Critical Dates Watchdog skill. Feed Claude a P&S, the skill will extract the critical dates, run a second pass to verify, and add them all to your calendar (after it checks in with you to confirm.) You can see how it works below:
Heads up, this skill is live in the Studio. Start your 7 day trial here to try it out along with the other skills we’ve purpose built for CRE.
But as much as we love skills, they aren’t always the right solution when you want to leverage AI. Especially if you want to automate the process.
A few nuances of using skills:
Skills tend to work best when you’re sitting next to them. Typically you’re telling AI to run the skill. That’s perfect if you’re spending all day working in Claude, especially if the skill is for a task where you’ll need to go back and forth with AI to edit the final output. But unless it’s something that can be scheduled based on a time/day of the week, skills aren’t the best fit for teeing up AI to help with a task automatically.
AI is driving the bus. When you call a skill, AI doesn't just follow your instructions, it interprets them, makes inferences, and decides where to go next. That's great when the job needs judgment, but it might be a problem when you need the same exact steps in long process run the same way every time, and it might be overkill when judgement isn’t needed much at all.
The costs add up: Every time you call a skill, AI re-reads and re-interprets the whole set of instructions from scratch, even the steps in the process that don't need any real thinking. You're paying for the model to think through the entire thing on every single run. On something you do once in a while, you'll never feel it. But if you want it running every day, or hundreds of times a month, it adds up, and you're paying full price for AI to think hard on the parts that don't need it.
The case for automated workflows:
Automation platforms, like Relay.App and Zapier, help you build step-by-step automated processes that can be triggered by things happening in external tools that you use everyday. And instead of having AI run the whole show, the workflow patches in AI for the steps of the process where it’s actually needed. For common, tedious, routine processes where you want the same output every time without AI making many decisions, these can often be the better route.
Confused? Here’s a few examples:
Underwriting: You have some deal documents, hand them to Claude, and run your underwriting skill. It analyzes the documents, pulls out the relevant intel, and builds out your underwriting model. This is a great skill. You need AI thinking through most of the process, you might not always have the same documents so it will have to use some judgement, and you’ll want to work with AI to finesse the model before it’s finished, so it’s fine if it only runs with you in the room.
Deal Tracking: You receive a new OM, AI extracts the information, runs a quick pass to see if it fits your buy-box, adds the intel to your active deal database, and sends your team an email with the deal summary. This is a great automation workflow. You don’t need to physically hand AI the OM, the workflow can be triggered by putting the OM in whatever platform you’d typically store it in. You don’t need AI to make judgement calls, or decide what’s relevant, you want to pull the same info and follow the same steps every time.
One involves judgement, iteration, and working back and forth with AI. The other can run on auto-pilot without you asking AI to do anything at all. Here’s one more difference: Not EVERYONE needs automated workflows, but everyone using AI should learn to build and use skills. Once you find something where AI helps you save time, you can likely save even more time by turning it into a skill. It’s a logical next step in customizing AI to work for you. But when you’re bumping up against a repetitive, tedious process where you don’t need AI running the whole show, challenge yourself to automate it.
Skills define how AI executes; markdown files store what it knows. That difference is its own lesson: CLAUDE.md Files vs. Claude Skills. And if you're still picking which Claude product to run skills in, start with Claude Chat vs. Cowork vs. Claude Code.